Getting The Best Performance With PySpark
-
Upload
spark-summit -
Category
Data & Analytics
-
view
1.997 -
download
4
Transcript of Getting The Best Performance With PySpark
![Page 1: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/1.jpg)
Getting the best Performance with PySpark
![Page 2: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/2.jpg)
Who am I?● My name is Holden Karau● Prefered pronouns are she/her● I’m a Principal Software Engineer at IBM’s Spark Technology Center● previously Alpine, Databricks, Google, Foursquare & Amazon● co-author of Learning Spark & Fast Data processing with Spark
○ co-author of a new book focused on Spark performance coming out next year*
● @holdenkarau● Slide share http://www.slideshare.net/hkarau ● Linkedin https://www.linkedin.com/in/holdenkarau ● Github https://github.com/holdenk ● Spark Videos http://bit.ly/holdenSparkVideos
![Page 3: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/3.jpg)
What is going to be covered:● What I think I might know about you● A quick background of how PySpark works● RDD re-use (caching, persistence levels, and checkpointing)● Working with key/value data
○ Why group key is evil and what we can do about it
● When Spark SQL can be amazing and wonderful● A brief introduction to Datasets (new in Spark 1.6)● Calling Scala code from Python with Spark
Torsten Reuschling
![Page 4: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/4.jpg)
Or….
Huang Yun Chung
![Page 5: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/5.jpg)
The different pieces of Spark
Apache Spark
SQL & DataFrames Streaming Language
APIs
Scala, Java, Python, & R
Graph Tools
Spark ML bagel & Graph X
MLLib Community Packages
Jon Ross
![Page 6: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/6.jpg)
Who I think you wonderful humans are?● Nice* people● Don’t mind pictures of cats● Know some Apache Spark● Want to scale your Apache Spark jobs● Don’t overly mind a grab-bag of topics
Lori Erickson
![Page 7: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/7.jpg)
A detour into PySpark’s internals
Photo by Bill Ward
![Page 8: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/8.jpg)
Spark in Scala, how does PySpark work?
● Py4J + pickling + magic○ This can be kind of slow sometimes
● RDDs are generally RDDs of pickled objects● Spark SQL (and DataFrames) avoid some of this
![Page 9: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/9.jpg)
So what does that look like?
Driver
py4j
Worker 1
Worker K
pipe
pipe
![Page 10: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/10.jpg)
So how does that impact PySpark?
● Data from Spark worker serialized and piped to Python worker○ Multiple iterator-to-iterator transformations are still pipelined :)
● Double serialization cost makes everything more expensive
● Python worker startup takes a bit of extra time● Python memory isn’t controlled by the JVM - easy to go
over container limits if deploying on YARN or similar● Error messages make ~0 sense● etc.
![Page 11: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/11.jpg)
Cat photo from http://galato901.deviantart.com/art/Cat-on-Work-Break-173043455
Photo from Cocoa Dream
![Page 12: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/12.jpg)
Lets look at some old stand bys:words = rdd.flatMap(lambda x: x.split(" "))
wordPairs = words.map(lambda w: (w, 1))
grouped = wordPairs.groupByKey()
grouped.mapValues(lambda counts: sum(counts))
warnings = rdd.filter(lambda x: x.lower.find("warning") != -1).
count()
Tomomi
![Page 13: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/13.jpg)
RDD re-use - sadly not magic● If we know we are going to re-use the RDD what should we do?
○ If it fits nicely in memory caching in memory○ persisting at another level
■ MEMORY, MEMORY_AND_DISK○ checkpointing
● Noisey clusters○ _2 & checkpointing can help
● persist first for checkpointing
Richard Gillin
![Page 14: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/14.jpg)
What is key skew and why do we care?● Keys aren’t evenly distributed
○ Sales by zip code, or records by city, etc.
● groupByKey will explode (but it's pretty easy to break)● We can have really unbalanced partitions
○ If we have enough key skew sortByKey could even fail○ Stragglers (uneven sharding can make some tasks take much longer)
Mitchell Joyce
![Page 15: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/15.jpg)
groupByKey - just how evil is it?● Pretty evil● Groups all of the records with the same key into a single record
○ Even if we immediately reduce it (e.g. sum it or similar)○ This can be too big to fit in memory, then our job fails
● Unless we are in SQL then happy pandas
PROgeckoam
![Page 16: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/16.jpg)
So what does that look like?(94110, A, B)(94110, A, C)(10003, D, E)(94110, E, F)
(94110, A, R)(10003, A, R)(94110, D, R)(94110, E, R)
(94110, E, R)(67843, T, R)(94110, T, R)(94110, T, R)
(67843, T, R)(10003, A, R)(94110, [(A, B), (A, C), (E, F), (A, R), (D, R), (E, R), (E, R), (T, R) (T, R)]
Tomomi
![Page 17: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/17.jpg)
“Normal” Word count w/RDDs
lines = sc.textFile(src)words = lines.flatMap(lambda x: x.split(" "))word_count = (words.map(lambda x: (x, 1)) .reduceByKey(lambda x, y: x+y))word_count.saveAsTextFile(output)
No data is read or processed until after this line
This is an “action” which forces spark to evaluate the RDD
These are still pipelined inside of the same python executor
![Page 18: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/18.jpg)
GroupByKey
![Page 19: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/19.jpg)
reduceByKey
![Page 20: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/20.jpg)
So what did we do instead?● reduceByKey
○ Works when the types are the same (e.g. in our summing version)
● aggregateByKey○ Doesn’t require the types to be the same (e.g. computing stats model or similar)
Allows Spark to pipeline the reduction & skip making the list
We also got a map-side reduction (note the difference in shuffled read)
![Page 21: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/21.jpg)
Can just the shuffle cause problems?
● Sorting by key can put all of the records in the same partition● We can run into partition size limits (around 2GB)● Or just get bad performance
● So we can handle data like the above we can add some “junk” to our key
(94110, A, B)(94110, A, C)(10003, D, E)(94110, E, F)
(94110, A, R)(10003, A, R)(94110, D, R)(94110, E, R)
(94110, E, R)(67843, T, R)(94110, T, R)(94110, T, R)
PROTodd Klassy
![Page 22: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/22.jpg)
Shuffle explosions :((94110, A, B)(94110, A, C)(10003, D, E)(94110, E, F)
(94110, A, R)(10003, A, R)(94110, D, R)(94110, E, R)
(94110, E, R)(67843, T, R)(94110, T, R)(94110, T, R)
(94110, A, B)(94110, A, C)(94110, E, F)(94110, A, R)(94110, D, R)(94110, E, R)(94110, E, R)(94110, T, R)(94110, T, R)
(67843, T, R)(10003, A, R)(10003, D, E)
javier_artiles
![Page 23: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/23.jpg)
100% less explosions(94110, A, B)(94110, A, C)(10003, D, E)(94110, E, F)
(94110, A, R)(10003, A, R)(94110, D, R)(94110, E, R)
(94110, E, R)(67843, T, R)(94110, T, R)(94110, T, R)
(94110_A, A, B)(94110_A, A, C)(94110_A, A, R)(94110_D, D, R)
(94110_T, T, R)(10003_A, A, R)(10003_D, D, E)(67843_T, T, R)
(94110_E, E, R)(94110_E, E, R)(94110_E, E, F)(94110_T, T, R)
Jennifer Williams
![Page 24: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/24.jpg)
Well there is a bit of magic in the shuffle….● We can reuse shuffle files● But it can (and does) explode*
Sculpture by Flaming Lotus GirlsPhoto by Zaskoda
![Page 25: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/25.jpg)
Our saviour from serialization: DataFrames
● For the most part keeps data in the JVM○ Notable exception is UDFs written in Python
● Takes our python calls and turns it into a query plan● If we need more than the native operations in Spark’s
DataFrames● be wary of Distributed Systems bringing claims of
usability….
![Page 26: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/26.jpg)
So what are Spark DataFrames?
● More than SQL tables● Not Pandas or R DataFrames● Semi-structured (have schema information)● tabular● work on expression instead of lambdas
○ e.g. df.filter(df.col(“happy”) == true) instead of rdd.filter(lambda x: x.happy == true))
![Page 27: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/27.jpg)
Where can Spark SQL benefit perf?● Structured or semi-structured data● OK with having less* complex operations available to us● We may only need to operate on a subset of the data
○ The fastest data to process isn’t even read
● Remember that non-magic cat? Its got some magic** now○ In part from peeking inside of boxes
● non-JVM (aka Python & R) users: saved from double serialization cost! :)
**Magic may cause stack overflow. Not valid in all states. Consult local magic bureau before attempting magic
Matti Mattila
![Page 28: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/28.jpg)
Why is Spark SQL good for those things?● Space efficient columnar cached representation● Able to push down operations to the data store● Optimizer is able to look inside of our operations
○ Regular spark can’t see inside our operations to spot the difference between (min(_, _)) and (append(_, _))
Matti Mattila
![Page 29: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/29.jpg)
How much faster can it be? (Scala)Andrew Skudder
![Page 30: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/30.jpg)
How much faster can it be? (Python)Andrew Skudder
*Note: do not compare absolute #s with previous graph - different dataset sizes because I forgot to write it down when I made the first one.
![Page 31: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/31.jpg)
Word count w/Dataframes
df = sqlCtx.read.load(src)# Returns an RDDwords = df.select("text").flatMap(lambda x: x.text.split(" "))words_df = words.map( lambda x: Row(word=x, cnt=1)).toDF()word_count = words_df.groupBy("word").sum()word_count.write.format("parquet").save("wc.parquet")
Still have the double serialization here :(
![Page 32: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/32.jpg)
Or we can make a UDF
def function(x):# Some magic
sqlContext.registerFunction(“name”, function, IntegerType())
![Page 33: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/33.jpg)
Buuuut….
● Our UDFs will be “slow” (e.g. require data copy from executor and back)
![Page 34: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/34.jpg)
Mixing Python & JVM code FTW:
● DataFrames are an example of pushing our processing to the JVM
● Python UDFS & maps lose this benefit● But we can write Scala UDFS and call them from
Python○ py4j error messages can be difficult to understand :(
● Trickier with RDDs since stores pickled objects
![Page 35: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/35.jpg)
Exposing functions to be callable from Python:
// functions we want to be callable from pythonobject functions { def kurtosis(e: Column): Column = new Column(Kurtosis(EvilSqlTools.getExpr(e))) def registerUdfs(sqlCtx: SQLContext): Unit = { sqlCtx.udf.register("rowKurtosis", helpers.rowKurtosis _) }}
![Page 36: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/36.jpg)
Calling the functions with py4j*:
● The SparkContext has a reference to the jvm (_jvm)● Many Python objects which are wrappers of JVM
objects have _j[objtype] to get the JVM object○ rdd._jrdd ○ df._jdf○ sc._jsc
● These are all private and may change
*The py4j bridge only exists on the driver**** Not exactly true but close enough
![Page 37: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/37.jpg)
e.g.:
def register_sql_extensions(sql_ctx): scala_sql_context = sql_ctx._ssql_ctx spark_ctx = sql_ctx._sc (spark_ctx._jvm.com.sparklingpandas.functions .registerUdfs(scala_sql_context))
![Page 38: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/38.jpg)
More things to keep in mind with DFs (in Python)
● Schema serialized as json from JVM● toPandas is essentially collect● joins can result in the cross product
○ big data x big data =~ out of memory● Pre 2.0: Use the HiveContext
○ you don’t need a hive install○ more powerful UDFs, window functions, etc.
![Page 39: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/39.jpg)
DataFrames aren’t quite as lazy...
● Keep track of schema information● Loading JSON data involves looking at the data● Before if we tried to load non-existent data wouldn’t fail
right away, now fails right away
Photo by Dan G
![Page 40: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/40.jpg)
The “future*”: Awesome UDFs
● Work going on in Scala land to translate simple Scala into SQL expressions - need the Dataset API○ Maybe we can try similar approaches with Python?
● Very early work going on to use Jython for simple UDFs (e.g. 2.7 compat & no native libraries) - SPARK-15369○ Early benchmarking w/word count 5% slower than native Scala UDF,
close to 65% faster than regular Python● Willing to share your Python UDFs for benchmarking? -
http://bit.ly/pySparkUDF *The future may or may not have better performance than today. But bun-bun the bunny has some lettuce so its ok!
![Page 41: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/41.jpg)
![Page 42: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/42.jpg)
The “future*”: Faster interchange
● Faster interchange between Python and Spark (e.g. Tungsten + Apache Arrow)? (SPARK-13391 & SPARK-13534)
● Willing to share your Python UDFs for benchmarking? - http://bit.ly/pySparkUDF
*The future may or may not have better performance than today. But bun-bun the bunny has some lettuce so its ok!
![Page 43: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/43.jpg)
Spark Testing Resources
● Libraries○ Scala: spark-testing-base (scalacheck & unit) sscheck (scalacheck)
example-spark (unit)○ Java: spark-testing-base (unit)○ Python: spark-testing-base (unittest2), pyspark.test (pytest)
● Strata San Jose Talk (up on YouTube)● Blog posts
○ Unit Testing Spark with Java by Jesse Anderson○ Making Apache Spark Testing Easy with Spark Testing Base○ Unit testing Apache Spark with py.test
raider of gin
![Page 44: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/44.jpg)
Additional Spark Resources
● Programming guide (along with JavaDoc, PyDoc, ScalaDoc, etc.)○ http://spark.apache.org/docs/latest/
● Kay Ousterhout’s work○ http://www.eecs.berkeley.edu/~keo/
● Books● Videos● Spark Office Hours
○ Normally in the bay area - will do Google Hangouts ones soon○ follow me on twitter for future ones - https://twitter.com/holdenkarau
raider of gin
![Page 45: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/45.jpg)
Learning Spark
Fast Data Processing with Spark(Out of Date)
Fast Data Processing with Spark (2nd edition)
Advanced Analytics with Spark
![Page 46: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/46.jpg)
Learning Spark
Fast Data Processing with Spark(Out of Date)
Fast Data Processing with Spark (2nd edition)
Advanced Analytics with Spark
Coming soon: Spark in Action
Coming soon:High Performance Spark
![Page 47: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/47.jpg)
And the next book…..
First four chapters are available in “Early Release”*:● Buy from O’Reilly - http://bit.ly/highPerfSpark● Chapter 9(ish) - Going Beyond ScalaGet notified when updated & finished:● http://www.highperformancespark.com ● https://twitter.com/highperfspark
* Early Release means extra mistakes, but also a chance to help us make a more awesome book.
![Page 48: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/48.jpg)
Spark Videos
● Apache Spark Youtube Channel● My Spark videos on YouTube -
○ http://bit.ly/holdenSparkVideos ● Spark Summit 2014 training● Paco’s Introduction to Apache Spark
![Page 49: Getting The Best Performance With PySpark](https://reader031.fdocuments.in/reader031/viewer/2022021813/586f7a261a28ab10258b720d/html5/thumbnails/49.jpg)
k thnx bye!
If you care about Spark testing and don’t hate surveys: http://bit.ly/holdenTestingSpark
Will tweet results “eventually” @holdenkarau
PySpark Users: Have some simple UDFs you wish ran faster you are willing to share?:http://bit.ly/pySparkUDF